Chapter 8 Nonlinear Analysis of Frames— An Introduction In this chapter we outline some of the ways in which the direct stiffness method of matrix analysis may be used in the nonlinear analysis of framed structures. The chapter starts with a brief discussion of nonlinear behavior, the aims of nonlinear analysis, and its place in structucal engineering. Sources of nonlinearity that are of common concern in frame design are listed, and the levels of analysis to be considered are defined. To provide perspective and a basis for comparison with the matrix methods, established Glassical solutions of some elementary problems are illustrated. Following this a sym bolic matrix stiffness method equation for each of the levels of analysis is presented Strategies for the numerical solution of nonlincar equations are outlined, but details of equation solution are left to Chapter 12. The chapter ends with a brief discussion of problems associated with formulating the equations of equilibrium on the deformed structuse, but a more thorough exploration of equation formulation is deferred to ‘Chapters 9 and 10. 8:1 NONLINEAR BEHAVIOR, ANALYSIS, AND DESIGN 216 ‘Most civil engineering structures behave in a linear elastic fashion under service loads. ‘Exceptions are slender structures such as some suspension systems, arches and tall buildings, and structures subject to early localized yielding or cracking. But prior to reaching their limit of resistance, almost all structures Would exhibit significant nonlin: tar response. Therefore, if linear elastic analysis is the highest level available, the design engineer must find another way to account for the effects that the analysis is incapable of simulating. The answer may lic in aay of the following: a) individwal judgment; b) code formulas that accept the results of a linear elastic or simpler analysis and make allowance for nonlinearity in some empirical or semi-empirical way; or ©) supplementary theoretical or experimental studies. In nonlinear analysis an attempt is made to improve the analytical simulation of the behavior of a structure in some respect. The fundamental aim is to improve the quality of design by providing the engineer witha more reliable prediction of the performance of a system that is under design or investigation. In making this closer link between, structural analysis and actual behavior, the traditional distinction between the terms “analysis” —the determination of forces and displacements under given loads—and “design””—the proportioning of members and connections to resist the determined ceficcis—becomes blurred (see Ref. 8.11). We will emphasize the analytical side of theBAL Nonlinear Behavior, Amalysis, and Des problem but the detennination of some aspect of behavior is the primary objective in the structures we consider. ‘Reflecting on the contents of the previous chaptors we can see that in linear elastic analysis, there is at in the reduction of the actual frame to a line diagram end in the assignment of boundary conditions and member propertics. But when tha: has been done, the results an anaytial problem that has a unique solution, the one that satisfies the requirements of compatibility and equilibrium on the undeformed structure, That js all that is meant by the term exact solution in reference to 8 Jinear analysis, Inthe two preceding chapters we have seen that there may be mathematical difficulties in the formulation ofthe linear equations and that we may have to settle for an approx~ jmation of that solution, Nevertheless, although there are many different approaches to linear elastic analysis, and although they may differ in mode end amouat of com. pulational effort, all methods that are capable of producing “an exact solation” rest fn the samne fo principles of mechanics, But the premise of linear elasti behavior forecioses the possibilty of revealing any manifestation of nonlinearity: A problem hnas been solved, but the solution may not tell us all we would lke to know xogarding the structure, Indeed, erucal information may be missing ‘Tn using noolinear analysis the uncertainty regarding actual behavior may be re- duced, In the process, however, the element of art in modeling the structe and in handling the equations of analysis is inereased. In modeling, the analyst must decide What sources of nonlinearity are apt to be significant and how fo represect them. In eating with the eesulting nonlinear equations, decisions must be made regarding how to reduce them to a system suitable for practicel computation and, then, tke method for solving the reduced system, in this and the folowing two chapters, procedures for the treatment of common situations will be presented. But the conditions encountered in practice are s0 vatied, and the devices for dealing with ther so many, that the coverage cannot be comprehensive 81.1 Sources of Nonlinearity In finear elastic analysis the material is assumed to be unyielding and its propesties invariable, and the equations of equilibrium are formulated on the geometey of the unloaded stinctre or, in the ease of self-strained structures, on an initial reference configuration (sce Section 5.3). Subsequent deformations are assumed to be so smell ts to be insignificant in their effect on the equilibrium and mode of response of the system. One consequence ofthis was our ability to treat axial force, bending moments, fand torques as uncoupled actions in developing the stiffness equations for elements of bisymmetrical section in Chapters 4 and 7. ‘Nonlinear analysis offers several options for addressing problems resulting from the above assumptions, We may consider only the geometric nonlinearity. Thatis, we may Continue to teeat the structural material as elastic but include the effects cf deforma- tions and nite displacements in formulating the equations of equilibrium. It is also possible to consider only the material nonlinearity, that i, the effect of changes in Inember material properties under load. And, as third general option, we may include tffects of both geometric and material nonlinearity in the analysis. In each case, the possibility of coupling of internal actions must be considered; it may be a dominant Feature of the analysis. Among the many sources of each class of nonlinearity are the following, Geometrical effects 1. Initial imperfections such as member camber and outof-plumb erection of a frame.248 Chapter 8 Nonlinear Analysis of Frames—An Introduetion 2. The P-A effect, a destabilizing moment equal to a gravity load times the horizon- tal displacement it undergoes as a result of the lateral displacement of the sup- porting structure, 43, ‘The P-d effect, the influence of axial force on the flexural stiffness of an individwal member? Material effects! 41. Plastic deformation of stec! structures, 2, Cracking or creep of reinforced eonerete structures. 3, Inclastie interaction of axial force, bending, shear, and torsion. Combined effects: 1. Plastic deformation plus P-A andlor P-6 effects. 2. Connection deformations. 3, Panel zone deformations. 4, Contributions of inflling and secondary systems to strength and stifiness. 84.2 Levels of Analysis Rarely ifever, is it possible to model al sources of nonlinearity and portray the actual bebavior of a practical structure in all of its detail. Normally, the problem is one of Ssleeting a method thet falls short of the ideal in onc way oF another but that does provide adequate analytical simlation of the case at hand. The most common levels Pr aanalyss are represented in Figure &.1 by schematic response curves for a statically Tonded frame. ‘The degree to which they can model true behavior differs, bt each can yield information of value to the engineer. ‘By defution, fst order (linear) elastic analysis excludes nonlinearity, but i genes ally represeats conditions at service loads very wel “The elastic critical oad is the load at which both the original and an alternative roading path become mathematically vai? and it can be shown that the path taken from that point wil be the alternative one. The load is commoaly determined from an tigenvatue analysis of an idealized elastic model of the structure”. The analysis also Silds the eigenvector, that is, the shape that the system assumes in the post-eitical Mite but it doesn’t determine its smplitude. The inelastic critica load is similarly de fined andcoteulated, but it isone in which the possibilty of precritialinelasticmaterial behavior is considered io the analysis, A critical load analysis may not include the nonlinear phenomena that must be taken jato account to determine accurately the hagnitude ofthe toad that would cause failure, but frequently, it gives adequate—and ‘aluable—pietures of possible types of failure : Tn second-order elastic analysis the offects of finite deformations and displacements ‘the Poh aod Petes desrtbed hee may be escouatered and treated as distinct phenomena so we define thems uch, Bu they often over in combination, jslying a blanket “Pela dofintion such a “the ree a Stet of col ace loads and ater] deietion onthe moxcatsin members (rom the Got fny of Referexce 8.) SiR sate is generally cefeed tao bifrctin of the Loading path 2p ts and other comparable methods of analysis it say be found that a mumber of values ofthe applied rradncury the titration comdion (eg. see Example 83), Exch may be called 2 “eral loud” Here, tec are reserving the term forthe smallest physialy meaning value ‘Polling whet bas become common in vil engineering structural cetinne pcre matheratial eene bu erly as shorten to devgaat ay method in whic the eters Fefered to re acount fo ysis, we ue the term second-ordersar Behavior, Analysis, and Design 219 lasicsnalyss aera 1080.3 ineleticsnaysie Lateral defection, & Figure 8. Levels of analysis, of the system are accounted for in formulating the equations of equilibrium. A second order elastic analysis can produce an excellent represcntation of destabilizing influ- tences stich as the P-A effect, bu it hasno provisions for detecting material nonlinearity. Several of the possible modes of naslinear elastic behavior are indicated in Figuee 8.1: (a) bifurestion (branching) of the loading path with the system following an alternative path in the post-eritcal state; (b) gradually increasing nonlinearity culminat ng in elas- tie instability at a limit point; and (¢) increasing stiffness either from th: onset of loading or, as shown, following 2 period of gradual softening. In first order inelastic analysis the equations of equilibrium arc written in termps of the geometry of the undeformed structure. Inelastic regions can develop gradually or, if the plastic hinge concept is adopted, as abrupt changes in the structure's response. When the destabilizing effects of finite displacements are relatively insignificant, first-order inelestic analysis can produce an excellent representation of simple clas- tic-plastic behavior and failure through mechanism formation, that is, the simple plastic lit load of Figure 8.1, But it has no provisions for detecting, geometric nonlinear effects and, of paramount concern, their influence oa the stability of the system. In second-order inelastic analysis the equations of equilibrium are written in ter of the geometry of the deformed system. It has the potential for accommodating all of the geometsc, elastic, and material factors that influence the response of estructure, ‘Thus, in principle —and in a deterministic sense—it enables the preparation of ana sal models capable of faithfully simalating aetual behavior and calculating the in- clastic stability limit, that is, the point at which a system’s capacity for resistance to additional load is exhausted and continued deformation results in a decrease in load- resisting capacity. “The path of increasingly nonlinear response (elastic or inelastic) culminating in in- stability is probably the most common mode of failure in civil engineering structures. For this reason, methods for detecting limit points will be a major concem ofthis text,81 Noullnear Behavior, Analysis, and Design 225 Member abis an ideally straight elastic bar. Summarize the small displacement theory of its flexural ? Behavior fo | La=o oy ae Davo 7 — 3. a # Oand P dvected upward L Ba 1. For a= 0. Proceeding as in Example 83, the same diferental equstion is obtained. The 1a P boundary conditions are: x = 0, v = O and ~ £, duldx = 0. Solving and satisfying the ‘boundary conditions results in slot we F t yr PEL I |: Fea = Te ® ' pif hk and the equation ofthe defected curve IND v= Gj sin © x 2, For a # 0, proceeding os in Example 83 leads tothe differential equation © and the general solution emul "The boundary conditions yield C; = Oand C; @ or, from Bq, (@), the deflection ©